Anthropic is developing its artificial intelligence systems at a rate twice as high as a year ago, despite the calls of CEO Dario Amodei to slow down the sector.
© Getty Images
05/11/2026
by
Lusa
A Anthropic is developing his artificial intelligence (IA) systems at a twice-upper rate a year ago, despite the appeals of his Executive President (CEO), Dario Amodei, for slowing down in the sector.
Anthropic invited researchers from the Model Evaluation and Threat Research (METR), based in Berkeley, California, to analyze their internal data and interview employees while preparing for their debut on Wall Street, in a context of heated debate about the possibility of AI models escaping human control.
METR estimated that Anthropic is improving the capabilities of its models to perform “well-specified tasks” at a pace about twice as fast as a year ago.
It was during the evaluation period of the research group that Anthropic CEO Dario Amodei called on the AI industry to slow down development so that security measures could keep pace. In an essay published on September 12, in response to the growing concerns about an uncontrolled AI, it urged the AI world to slow down the pace at which model capabilities improve.
METR has been carrying out capacity assessments of Anthropic, OpenAI and other laboratories for several years. The non-profit organisation is funded through donations and does not accept remuneration for its work, but the AI companies with which it has established partnerships for evaluations have provided it with access and tokens, which METR uses for evaluations, research and engineering.
One of its two reports was based on Anthropic's internal data analysis provided by the company over a period of about 11 weeks, from mid-August to the end of October, as well as on employee interviews.